Skip to content

Latest commit

 

History

History
246 lines (173 loc) · 13.1 KB

README.md

File metadata and controls

246 lines (173 loc) · 13.1 KB

SNT

The ImageJ framework for quantification of neuronal anatomy

SNT is ImageJ's framework for semi-automated tracing, visualization, quantitative analyses and modeling of neuronal morphology. For tracing, SNT supports modern multidimensional microscopy data, and highly-customizable routines. For data analysis, SNT features advanced visualization tools, access to all major morphology databases, and support for whole-brain circuitry data.

SNT can be used as a regular application or as a scripting library. Python (through pyimagej) and all of SciJava's scripting languages are supported. It is distributed with Fiji and supersedes the original Simple Neurite Tracer plug-in. It also incorporates several other neuroanatomy-related Fiji plugins. See SNT's publication and techical notes for details.

Overview

Overview

Features

For an overview of SNT capabilities have a look at the showcase gallery.

Detailed List

Tracing

  • Semi-automated Tracing:

    • Support for up to 5D multidimensional images, including multichannel, and timelapse sequences
    • Support for both ImageJ1 and ImgLib2 data structures
    • Several bidirectional search algorithms (A*, NBA*, Fast marching) with adjustable cost functions allow for efficient computation of curvatures for a wide range of imagery, that are up to 20x faster relatively to the original Simple Neurite Tracer plugin
    • Tracing in "secondary layers". This allows for paths to be computed on "enhanced" (pre-processed) images while interacting with the unfiltered, original image (or vice-versa). Toggling between the two data sources is immediate
    • Precise placement of nodes is aided by a local search that automatically snaps the cursor to neurites wihin a 3D neighborhood
  • Auto-tracing:

    • Generation of traces from thresholded/filtered images
    • Machine learning: Built-in routines for training random forest classifiers on previously traced paths (LabKit/Trainable Weka segmentation bridges)
  • Tracing can be interleaved with image processing routines

  • Tracing is scriptable. Interactive scripts allow for real-time inspection of results

  • Paths can be tagged, searched, grouped and filtered by morphometric properties (length, radius, etc.)

  • Paths can be edited, i.e., a path can be linked or merged together, or split into two. Nodes can be moved, deleted, or inserted

  • Post-hoc refinement of node positioning and radii by 'fitting' traces to the fluorescent signal associated with a path

Analysis

  • Extensive repertoire of metrics, namely those provided by L-measure and NeuroM. Metrics can be collected from groups of cells, single cells, or parts thereof

  • Analysis based on neuropil annotations for whole-brain data such as MouseLight

  • Direct access to public databases, including FlyCircuit, Insect Brain Database, MouseLight, NeuroMorpho, and Virtual Fly Brain

  • Built-in commands for immediate retrieval of statistical reports, including summary statistics, tests (two-sample t-test/one-way ANOVA), comparison plots and histograms

  • Convex hull analyses

  • Graph theory-based analyses

  • Persistent homology-based analyses

  • Sholl and Horton-Strahler analyses

  • Image processing workflows: Reconstructions can be converted to masks and ROIs. Voxel intensities can be profiled around (or across) traced paths

Visualization

  • Quantitative visualizations: Display neurons color coded by morphometric traits, or neuropil annotations.

  • Publication-quality visualizations: Neuronal reconstructions, diagrams, plots and histograms can be exported as vector graphics

  • Reconstruction Viewer: Standalone hardware-accelerated 3D visualization tool for both meshes and reconstructions.

    • Interactive and programmatic scenes (controlled rotations, panning, zoom, scaling, animation, "dark/light mode", etc.)
    • Customizable views: Interactive management of scene elements, controls for transparency, color interpolation, lightning, path smoothing, etc. Ability to render both local and remote files on the same scene
    • Built-in support for several template brains: Drosophila, zebrafish, and Allen CCF (Allen Mouse Brain Atlas)
  • sciview integration

  • Graph Viewer: A dedicated viewer for graph-theory-based diagrams

    • Display reconstructions as dendrograms
    • Quantitative connectivity graphs for single cells and groups of cells

Scripting

  • Almost every aspect of the program can be scripted in any of the IJ2 supported languages, or from Python through pyimagej

  • Detailed documentation and examples, including Python notebooks, and end-to-end examples

  • Headless scripts supported

  • (Experimental) Script Recorder

Modeling

  • Biophysical modeling of neuronal growth is performed through Cortex3D (Cx3D) and sciview, in which a modified version of Cx3D grows neuronal processes with sciview’s data structures.

Compatibility

  • Support for multiple file formats including SWC, TRACES, JSON (MouseLight specification), and NDF (NeuronJ data file)

  • Backwards compatibility: Special effort was put into backwards compatibility with Simple Neurite Tracer (including TrakEM2 and ITK interaction)

  • Aggregation of legacy plugins

Installation

SNT is a Fiji plugin, currently distributed through the Neuroanatomy update site.

The first time you start SNT from Fiji's menu structure (Plugins>Neuroanatomy>SNT...) you should be prompted for automatic subscription and download of required dependencies. If not:

  1. Run the Fiji Updater (Help › Update..., the penultimate entry in the Help › menu)
  2. Click Manage update sites
  3. Select the Neuroanatomy checkbox
  4. Optionally, select the sciview checkbox. This is only required for extra sciview functionality
  5. Click Apply changes and Restart Fiji. SNT commands are registered under Plugins>Neuroanatomy> in the main menu and SNT scripts under Templates>Neuroanatomy> in Fiji's Script Editor.

Problems? Have a look at the full documentation.

Developing

On the cloud

Use this button to open the project on the cloud using Gitpod. No local installation necessary (although project may take a while to load).

Open in Gitpod

Locally

  1. Clone the main branch of this repository (use the green code button above the list of files)

  2. Import the project into an IDE such as Eclipse/IntelliJ/NetBeans:

    • In Eclipse: Run Import> Existing Maven Projects and specify the path to the downloaded SNT folder in Root Directory
    • In IntelliJ: In the Welcome Prompt, choose Open or Import and specify the path to the downloaded SNT folder
    • In NetBeans: Run File> Open Project..., select the downloaded SNT directory, and click on Open Project
  3. Wait for all the dependencies to be downloaded, and locate the StartImageJAndSNTDemo class in the tests folder.

  4. Java 21 is recommended to run SNT, so you should specify it as the project JDK. However using Java17+ or newer requires the following VM arguments to be specified: --add-opens java.base/java.lang=ALL-UNNAMED. To do so:

    • In Eclipse: Run -> Run Configurations..., Arguments tab
    • In IntelliJ: Run -> Edit Configurations..., Add VM Options (Alt+V)
  5. Run SciviewSNTDemo.main()

Useful Resources to Start Hacking SNT

From a Java IDE:

  • Test demos
  • main() methods found on most classes: Frequently, these showcase the class's functionality
  • JUnit tests

From Fiji's Script Editor:

  • Scripts in the Templates>Neuroanatomy> menu. These are part of the source code and can also be accessed from Script templates

From python:

Snippets and code templates:

Contributing

Want to contribute? Please, please do! We welcome issues and pull requests any time. You can also report bugs and propose improvements using the forum. Please tag your post using snt so that it does not go unnoticed.

Thanks To All Contributors

Thanks a lot for spending your time helping SNT!

Contributors